Special Issue / Special Column

Using artificial intelligence in airspace system to improve airspace hierarch...

CHEN Zhijie, TANG Jinhui, WANG Chong, CHENG Jizeng, CAO Shan, SHAO Xin

doi:10.7527/S1000-6893.2020.25018

2021, (4): 525018-525018.

Abstract

To follow the development trend of unmanned platform, diversified users and personalized services in the future aviation field, hierarchical governance is being adopted in airspace operation. With the continuous improvement and accumulation of computing power, algorithms and data, the data driven artificial intelligence method will continue to empower hierarchical airspace system. Firstly, this paper combs the development trend of China's airspace system in terms of five hierarchical scenarios: ultra-low altitude transportation, urban transportation, regional transportation, hub transportation and suborbital transportation. The core difficulties and key issues of airspace operation are summarized. Secondly, the research framework, research contents and key technologies of data-driven artificial intelligence method to solve the scientific problems of airspace operation are proposed. Specific cases of application of artificial intelligence for hierarchical scenarios of airspace are brief...

Thoughts on technology development of OODA empowered with AI

ZHU Xuejun, ZHAO Changjian, LIANG Zhuo, TAN Qingke

doi:10.7527/S1000-6893.2020.24332

2021, (4): 524332-524332.

Abstract

With the rapid development of science and technology, system combat supported by informationization will be a major style of the future war. "Winning by faster based on Observation-Orientation-Decision-Action (OODA)" has become an important mechanism for victory in modern wars. As the battlefield environment is going more and more complex and the confrontation is extending into variable domains and dimensions, the mapping relationship from battlefield situation to combat strategy becomes extremely complicated, posing new challenges to the OODA loop. To ensure that the OODA loop meets the task requirements, the OODA loop theory is combined with Artifical Intelligence(AI) to drive efficient operation of each loop and shorten the OODA loop time, so as to provide key support for winning the war. This paper summarizes the progress of application of AI in the military field, and analyzes the challenges of OODA empowered with AI. Then, preliminary thoughts on systematic development of the rel...

Prospect of artificial intelligence empowered fluid mechanics

ZHANG Weiwei, KOU Jiaqing, LIU Yilang

doi:10.7527/S1000-6893.2020.24689

2021, (4): 524689-524689.

Abstract

Artificial Intelligence (AI) is an advanced technology in the 21 st century. For researchers in related fields, rejuvenation of fluid mechanics in the age of intelligence is worth consideration. This paper proposes intelligence empowered fluid mechanics, explaining and summarizing its meaning, important topics, research progress, and research difficulties. The future development of intelligent fluid mechanics is also discussed. The research points out that the data generated in computational fluid dynamics or experiments are inherently big data, and how to use these data through machine learning methods, like the deep neural network, random forest, and reinforcement learning, to alleviate or even replace the dependence on human brain is a new research paradigm; at the theoretical and methodological level, main research topics cover machine learning of the governing equations and turbulence modeling, the intellectualization of dimensional and scaling analysis, as well as numerical simul...

State estimation method for spacecraft autonomous navigation: Review

WANG Dayi, HOU Bowen, WANG Jiongqi, GE Dongming, LI Maodeng, XU Chao, ZHOU Haiyin

doi:10.7527/S1000-6893.2020.24310

2021, (4): 524310-524310.

Abstract

Autonomous navigation is the key technology of spacecraft autonomous operation, while state estimation, the core means of spacecraft autonomous navigation which refers to the real-time determination of spacecraft orbit position, velocity, attitude and other navigation parameters, is one of the key development directions of the spacecraft autonomous navigation technology. Aiming at the practical requirements of the spacecraft autonomous navigation, this paper illustrates the necessity of studying spacecraft autonomous navigation state estimation method and introduces its research status from three aspects, including observability analysis of the navigation system, the navigation filtering algorithm, and error compensation of the navigation system. Then practical applications of the state estimation method in the spacecraft autonomous navigation system is analyzed and summarized. Finally, based on theoretical research and practical applications, the main problems of the state estimation ...

Research status and prospect of distributed optimization for multiple aircraf...

JIANG Xia, ZENG Xianlin, SUN Jian, CHEN Jie

doi:10.7527/S1000-6893.2020.24551

2021, (4): 524551-524551.

Abstract

In the aviation field, practical tasks such as coordinated search and rescue, area monitoring, and formation control of multiple aircraft are conducted by many individuals with distributed information and complex task objectives. Distributed optimization is an important guarantee for the effective coordination of multiple aircraft in the above tasks, having significant theoretical and practical value. A brief introduction to classical distributed optimization tasks in the field of aviation is presented, and a research overview of distributed optimization work is conducted from three perspectives: problem models, research frameworks and classical algorithms of optimization problems. According to the optimization problems, the classical research works in the field of distributed optimization are summarized from four aspects: unconstrained distributed optimization, distributed optimization with set constraints, distributed optimization with inequality constraints, and distributed non-conv...

Spacecraft guidance and control based on artificial intelligence: Review

HUANG Xuxing, LI Shuang, YANG Bin, SUN Pan, LIU Xuewen, LIU Xinyan

doi:10.7527/S1000-6893.2020.24201

2021, (4): 524201-524201.

Abstract

Spacecraft guidance and control technology is one of the key technologies to ensure successful implementation of space missions. Currently, the strong nonlinearity of dynamic models and the uncertainty of parameters restrict the development of high-precision attitude and orbit control technology, while the system failure determines the success or failure of the spacecraft attitude and orbit control. The new generation of artificial intelligence technology represented by machine learning shows tremendous potential in the field of spacecraft guidance and control. This paper first summarizes the research development and application status of trajectory guidance and attitude control based on artificial intelligence technology, and analyzes the development trend of spacecraft trajectory planning, attitude control, fault diagnosis and fault-tolerant control technology. Then, from the four aspects of robust trajectory planning, adaptive attitude control, rapid fault diagnosis, and adaptive fa...

A review on development of intelligent health management technology for space...

YUAN Li, WANG Shuyi

doi:10.7527/S1000-6893.2020.25044

2021, (4): 525044-525044.

Abstract

As one of the key technologies for spacecraft intelligent autonomous control, health management is an effective way to improve the security, reliability and stability of spacecraft. Based on the development trend of artificial intelligence technology and the new general architecture of spacecraft intelligent autonomous control system that is developed by our team, this paper gives a review of the status and development trend of intelligent health management technology for spacecraft control system. First, the challenges of health management technology for spacecraft control systems in the process of design, test and in-orbit operation are presented. Then, the states of the art of the health management technology based on artificial intelligence and its applications in the aerospace field are discussed in terms of fault prognosis, fault diagnosis and life assessment. Finally, possible development directions of the health management technology for spacecraft control system are summarized.

Object detection in UAV imagery based on deep learning: Review

JIANG Bo, QU Ruokun, LI Yandong, LI Chenglong

doi:10.7527/S1000-6893.2020.24519

2021, (4): 524519-524519.

Abstract

Object detection is one of the key technologies in improving the autonomous sensing ability of Unmanned Aerial Vehicles (UAVs). Research on object detection is of critical significance in UAV applications. Compared with traditional methods based on manual features, deep learning based on the convolutional neural network has a powerful capability of feature learning and expression, therefore becoming the mainstream algorithm in object detection. In recent years, object detection research has achieved a series breakthrough in the field of natural scene and the research in UAVs has increasingly become a hotspot simultaneously. This paper reviews the research progress of object detection algorithms based on deep learning, summarizing their advantages and disadvantages. Then, some typical aerial image datasets and the method of transfer learning are introduced, and relevant algorithms are analyzed aiming at the complex background, small and rotating objects, large fields of view in UAV imag...

Machine learning for flow control: Applications and development trends

REN Feng, GAO Chuanqiang, TANG Hui

doi:10.7527/S1000-6893.2020.24686

2021, (4): 524686-524686.

Abstract

As a multidisciplinary field in fluid mechanics, flow control has played a key role in both scientific research and engineering applications. Due to complicated features of flow systems such as strong nonlinearity, flow control, especially closed-loop control, has been a challenging issue in the past decades. Recently, the rapid developing machine learning has brought new methods, new perspectives, and new views to diverse fields, and also to flow control. This article reviews three distinct ideas that involve machine learning into flow control, so as to demonstrate an overall view of machine learning in flow control, and furthermore, to outline some trends for this field.

Turbulent flow data assimilation and its applications

HE Chuangxin, DENG Zhiwen, LIU Yingzheng

doi:10.7527/S1000-6893.2020.24704

2021, (4): 524704-524704.

Abstract

Data Assimilation (DA) has been introduced into the turbulence dynamics community in recent years. Coupling experimental measurements and numerical simulation, it improves the accuracy and scope of measurements and reduces the uncertainty of simulations. Experimental observations, predictive models and assimilation algorithms are three essential factors in DA. Observations in turbulent flows usually involve hot wire anemometer, Particle Image Velocimetry (PIV), pressure sensors and other measurement techniques. The predictive model refers specifically to flow governing equations and turbulence closures. The assimilation algorithm ranges from Bayesian inference, Ensemble Karman Filter (EnKF), to adjoint formulations. DA for steady-state flows has a combination of Reynolds-Averaged Navier-Stokes (RANS) turbulence models, aiming at model constant recalibration, equation form-error correction and Reynolds stress term reproduction, whilst the unsteady DA has two main categories, i.e., four-...

Progress in deep convolutional neural network based flow field recognition an...

YE Shuran, ZHANG Zhen, WANG Yiwei, HUANG Chenguang

doi:10.7527/S1000-6893.2020.24736

2021, (4): 524736-524736.

Abstract

With excellent performance, the deep learning architecture has enabled new developments in application of machine learning in fluid mechanics, and can cope with many challenges and needs in fluid mechanics. Due to powerful nonlinear mapping capabilities and hierarchical extraction of information features, the Convolutional Neural Network (CNN) has become a tool that cannot be ignored in current research on flow features. This paper summarizes the progress and achievements in this research area. First, the developments of deep learning for fluid mechanics and CNNs are briefly reviewed. Then, the research progress of using deep CNN in flow prediction, flow shape optimization, improving the accuracy of flow field visualization, and generation confrontation is introduced. Finally, prospects of application of deep learning in flow field recognition are discussed to provide a reference for subsequent research.

Discussion on technologies for intelligent spectrum management and control un...

DING Guoru, SUN Jiachen, WANG Haichao, JIAO Yutao

doi:10.7527/S1000-6893.2020.24750

2021, (4): 524750-524750.

Abstract

As an important component of national land space, the electromagnetic spectrum space presents many new challenges, such as the intricacy and complexity of the electromagnetic environment, diversity of electromagnetic targets and variability of spectrum usage behaviors. As a result, spectrum security becomes an increasingly urgent issue. For the common national defense requirements of security for spectrum order, security for spectrum confrontation and security for spectrum sharing in the complex electromagnetic environment, spectrum management and control based on artificial intelligence has become the most important research orientation in the radio spectrum field. There exist the challenging fundamental theoretical and technical problems in this research field. This paper firstly investigates the national strategic demands for intelligent spectrum management and control under complex electromagnetic environments, and then summarizes the significance and technical challenges of intell...

Aircraft intelligent design: Visions and key technologies

LI Ni, BU Shuhui, SHANG Bolin, LI Yongbo, TANG Zhili, ZHANG Weiwei

doi:10.7527/S1000-6893.2020.24752

2021, (4): 524752-524752.

Abstract

Nowadays, the aircraft has been more and more autonomous and intelligent, and their missions have been more complex. Various types of aircraft have been developing, such as hyper sonic aircraft, highly stealth aircraft, and morphing aircraft. Thus, to fully achieve the comprehensive performance of the aircraft, an intelligent and multidisciplinary-integrated design process needs to be developed to improve the traditional decoupled design process. In this paper, the problems existing in the traditional aircraft design process are firstly discussed. Then, an intelligent and full-life-cycle aircraft design framework is proposed. In this closed-loop aircraft design framework, the knowledge base is applied to connect the stages of design, manufacturing, and operation and maintenance. The intelligent digital twin technology is employed to simulate, analyze and predict the states of aircraft, so as to update the data needed in the process of aircraft design and operation. The key technologies...

Progress of deep learning modeling technology for fluid mechanics

WANG Yixing, HAN Renkun, LIU Ziyang, ZHANG Yang, CHEN Gang

doi:10.7527/S1000-6893.2020.24779

2021, (4): 524779-524779.

Abstract

Deep learning technology has brought subversive changes in many fields, such as image processing, language translation, disease diagnosis, and game competition. Due to the characteristics of high dimensionality, strong nonlinearity and large amount of data, fluid mechanics is an important area where deep learning is good at and could bring out innovation in research paradigm. At present, the deep learning technology has been initially applied in the field of fluid mechanics, and its application potential has been gradually confirmed. Based on the deep learning technology for fluid mechanics and the recent research results of our group, this paper discusses the deep learning modeling technology for fluid mechanics and its latest progress. First, the basic theory of the deep learning technology is introduced, and the mathematics behind the deep learning methods commonly used in fluid mechanics modeling are explained. Then, the progress of the deep learning technology involved in several ...

Review on basic concept and applications for artificial intelligence in aviat...

LU Xinlai, DU Ziliang, XU Yun

doi:10.7527/S1000-6893.2021.25150

2021, (4): 525150-525150.

Abstract

Recent years have witnessed rapid development of artificial intelligence application in aviation industry. In order to illustrate some basic but fundamental concepts, the definition of artificial intelligence and intelligence levels are discussed. Experts generally agree that it will not be a commonly accepted definition of AI in decades, and the industrial community should deal with the certainty and uncertainty of state of art artificial intelligence with dialectical perspectives. For intelligence systems that execute specific tasks in certain military context, classifying their intelligence level is unnecessary. From the perspective of general history of development, airborne missiles, airborne system and trustworthiness, the characteristics and trends of artificial intelligence adoption in aviation industry are illustrated. The importance of trustworthiness of artificial intelligence in aviation as a precondition for better adoption is emphasized.

Reviews on intelligent flight technology of civil aircraft

YANG Zhigang, ZHANG Jiong, LI Bo, ZENG Rui, MAO Yanxun

doi:10.7527/S1000-6893.2020.25198

2021, (4): 525198-525198.

Abstract

Artificial Intelligence (AI) has become a new competing track to innovation, development, and competition of the civil aircraft technology. In a variety of applications of AI for the full life cycle of civil aircraft, intelligent flight is committed to changing the traditional flight driving mode and reconstructing the human-computer interaction mode and air traffic management architecture of future flight, and has become the most significant and subversively innovative features of the industry and the new arena of the civil aircraft intelligent competition. This article enunciates the basic concepts of intelligent flight, and its three-phase roadmap: the auxiliary intelligent, enhanced intelligence and complete intelligence. The planning strategy for the intelligence flight and relevant technical systems in accordance with the CCAR 23&25 are constructed, and the light sport aircraft requirements are analyzed. The demanding technologies (reliable airworthiness of AI, human factor engin...

Reinforcement learning method for supercritical airfoil aerodynamic design

LI Runze, ZHANG Yufei, CHEN Haixin

doi:10.7527/S1000-6893.2020.23810

2021, (4): 523810-523810.

Abstract

Reinforcement learning as a machine learning method for learning policies learns in a way similar to human learning process, interacting with the environment and learning how to achieve more rewards. The elements and algorithms of reinforcement learning are defined and adjusted in this paper for the supercritical airfoil aerodynamic design process. The results of imitation learning are then studied, and the policies from the imitation learning are adopted in reinforcement learning. The influence of different pretraining processes is studied, and the final policies tested in other similar environments. The results show that pretraining can improve reinforcement learning efficiency and policy robustness. The final policies obtained in this study can also have satisfactory performance in other similar environments.

Missile aerodynamic performance prediction of Gaussian process through automa...

HU Weijie, HUANG Zenghui, LIU Xuejun, LYU Hongqiang

doi:10.7527/S1000-6893.2020.24093

2021, (4): 524093-524093.

Abstract

The initial stage of missile design usually requires quick and rough evaluation of missile aerodynamic performance. To improve the low calculation accuracy of traditional engineering estimation software and reduce the high calculation cost of CFD method, we propose a scheme based on the Gaussian Process Regression (GPR) surrogate model to quickly and accurately predict the aerodynamic performance of typical missiles. The prediction results of the GPR model are analyzed, taking the missile shape parameters and angle of attack as the input and the lift coefficient, drag coefficient and moment coefficient as the output. First of all, compared with the prediction accuracy of other commonly used surrogate models, the prediction errors of the GPR model for the three coefficients are only 0.24%, 0.36% and 0.36%, respectively, lower than those of other surrogate models. Secondly, considering the problem that the kernel function selection of the GPR model depends heavily on artificial prior kno...

Aeroheating agent model based on radial basis function neural network

ZHANG Zhichao, GAO Taiyuan, ZHANG Lei, TUO Shuangfen

doi:10.7527/S1000-6893.2020.24167

2021, (4): 524167-524167.

Abstract

An aeroheating agent model based on the radial basis function neural network is proposed to rapidly acquire heat flux on the surface of hypersonic vehicles and shorten the aeroheating design cycle. A regularization radial basis function neural network is firstly constructed on each grid node of the solid surface, followed by the acquisition of the connection weights of different nerve cells through simultaneous training of all neural networks based on the data of the train set. Finally, the heat flux results of different positions on the surface of vehicles are predicted synergistically by the neural networks on grid nodes. The simulation results of the elliptical blunt vehicle designed by the Mars Science Laboratory of NASA indicate that the agent model can be employed to rapidly predict heat flux of hypersonic vehicles with good generalization capability. In addition, the results of heat flux on the stagnation point and the windward wall prove that the heat flux deviations between th...

Prediction of single-row hole film cooling performance based on deep learning

LI Zuobiao, WEN Fengbo, TANG Xiaolei, SU Liangjun, WANG Songtao

doi:10.7527/S1000-6893.2020.24331

2021, (4): 524331-524331.

Abstract

Film cooling is an effective way to enhance the high temperature resistance of turbine blades and indirectly increase the inlet temperature of the turbine. Currently, the mainstream design method of the film cooling hole layout is to select and optimize the initial schemes with Computational Fluid Dynamics (CFD), followed by model experiments. However, this method has a long design period and thus is time-consuming. The traditional empirical formula method for rapid evaluation of cooling efficiency is problematic in that the function form is complex, the fitting precision is low and the applicable range of parameters is narrow. This paper designs a deep neural network with a Multi-Layer Perceptron model (MLP) based on the principle of deep learning, and establishes a prediction model for adiabatic film cooling efficiency. The calculation of CFD is utilized to train the network, and the results obtained from the modeling efforts indicate that the deep learning model has a fitting degree...

Machine learning method for correction of rarefied nonlinear constitutive rel...

LI Tingwei, ZHANG Mang, ZHAO Wenwen, CHEN Weifang, JIANG Lijian

doi:10.7527/S1000-6893.2020.24386

2021, (4): 524386-524386.

Abstract

The continuum medium hypothesis in the rarefied non-equilibrium flow field has been invalid, and the rarefied non-equilibrium flow is mainly researched around the Boltzmann equation with the Unified Gas-Kinetic Scheme (UGKS) as a representative method. In numerical simulation of the rarefied non-equilibrium flow, the Navier-Stokes (N-S) equation has high efficiency but low accuracy while the UGKS method has high accuracy but low efficiency. In this paper, a Data-driven method for the solution of the Nonlinear Constitutive Relations of the rarefied non-equilibrium flow based on the N-S equation and the UGKS method (DNCR) is proposed. The flow field numerical simulation results of the N-S solver and the UGKS solver are used as the data set. Based on the characteristic parameters of the flow field, an extremely randomized trees algorithm is adopted to nonlinearly correct the linear viscous stress term and heat flux term of the N-S equation. The numerical solution of the rarefied non-equil...

Application of field inversion based on discrete adjoint method in turbulence...

YAN Chongyang, ZHANG Yufei, CHEN Haixin

doi:10.7527/S1000-6893.2020.24695

2021, (4): 524695-524695.

Abstract

Accurate simulation of turbulent flow is a common problem in engineering and academic fields. In this paper, the idea of data-driven turbulence modeling is adopted, and a framework of flow field inversion based on discrete adjoint is established. The SA model is modified by multiplying the production term of its eddy viscosity transport equation and a coefficient with non-uniform distribution, which is inferred with limited observation data. To improve the efficiency of discrete adjoint optimization under physical constraints, the constraint-augmented adjoint method is used, and its efficiency is verified in this paper. Two cases of iced airfoil and periodic hill are selected for analysis. The results obtained in both cases are highly consistent with the observed data, and the limited observation information can be generalized to the whole flow field with the help of the correction of the turbulence model. The analysis shows that the correction region deduced from field inversion has a...

Knowledge discovery for vehicle aerodynamic configuration design using data m...

LIU Shenshen, Chen Jiangtao, GUI Yewei, TANG Wei, WANG Anling, HAN Qinghua

doi:10.7527/S1000-6893.2020.24708

2021, (4): 524708-524708.

Abstract

To gain a deeper understanding of the relationship between multiple objectives and multiple design parameters in the optimization process of vehicle aerodynamic configuration design and improve the scientificity and efficiency of the optimization model, we study the knowledge discovery of aircraft aerodynamic configuration design based on data mining methods. Four machine learning methods including analysis of variance, decision tree, isometric mapping, and self-organizing map are applied to data mining for aerodynamic design space of a hypersonic glide vehicle configuration optimization problem. Trade-offs between four objective functions (lift-to-drag ratio, lateral/side stability and volumetric efficiency) and influences of the design variables on the objective functions obtained quantitatively and qualitatively by the four methods are presented and discussed. Meanwhile, the design rules for variable values to generate better results are also analyzed. The features of the four data ...

Spatial calibration model of stereo PIV based on neural network

DOU Jianyu, PAN Chong

doi:10.7527/S1000-6893.2020.24720

2021, (4): 524720-524720.

Abstract

The accuracy of spatial calibration in Stereo Particle Image Velocimetry (SPIV) has a considerable influence on the accuracy of velocity measurement. To study the ability of various calibration models to handle input errors, we define a dimensionless parameter, namely, the error attenuation coefficient, to evaluate the response of spatial calibration models to input errors. Based on this error attenuation coefficient, the error propagation characteristics of conventional spatial calibration models, including the polynomial model and the camera pinhole model, can be quantitatively evaluated. A neural network-based space calibration model is then developed. Unlike conventional calibration models, this new model is naturally adaptive to multiple-camera joint calibration, thus suitable for SPIV. Synthetic experiments demonstrate the ability of this neural network model to suppress the propagation of the input error in a large measurement parameter space, which is not possessed by the polyn...

Prediction of CMG failure boundary domain based on combined stress test and n...

HUANG Shouqing, LIU Shouwen, ZHAI Baichen, ZHOU Yuan, HUANG Xiaokai, QIN Taichun

doi:10.7527/S1000-6893.2020.24208

2021, (4): 524208-524208.

Abstract

This study designs the test equipment for simultaneous simulation of the vacuum thermal environment and angular momentum exchange conditions between a CMG (Control Moment Gyroscope) and a spacecraft, and proposes a combined stress working-state test method to simulate the in orbit vacuum temperature, CMG gimbal rotational speed and spacecraft rotational speed. A quantitative expression method for CMG running status suitable for a neural network model is applied. Based on relatively little amount of test results, the neural network model after training is used to predict the working limit rotational speed matrix, the failure boundary and the failure boundary domain. Furthermore, the influence of experience samples on the prediction results is analyzed, and the coupling effect of each stress on the working domain of other stresses investigated, and the predicted initial value by the neural network proposed to reflect the reliability of the prediction results. The results show that the pr...

Sensitivity analysis of key design parameters of commercial aircraft using de...

FAN Zhouwei, YU Xiongqing, WANG Chao, ZHONG Bowen

doi:10.7527/S1000-6893.2020.24353

2021, (4): 524353-524353.

Abstract

The sensitivity analysis of key design parameters of aircraft reveals the relationship between the key design parameters and aircraft characteristics, facilitating the decision making in aircraft preliminary design. Aiming at the key design parameter sensitivity of wide-body commercial aircraft, we establish a deep neural network model based on the features of key design parameters and aircraft characteristics and the coupling relationship among multiple disciplines, taking the key design parameters as input to predict the aircraft characteristics. In this model, multiple input layers, multiple output layers, and multiple blocks of hidden layers are set to simulate the effects of key design parameters on aircraft characteristics and the interactions among different aircraft characteristics. Comparisons with traditional surrogate models reveal that the deep neural network model has higher prediction accuracy and better adaptability to the aircraft characteristics. The proposed model is ...

Operational reliability of aircraft power plant based on machine learning

FENG Yunwen, PAN Weihuang, LIU Jiaqi, LU Cheng, XUE Xiaofeng, LENG Jiaxing

doi:10.7527/S1000-6893.2020.24732

2021, (4): 524732-524732.

Abstract

To study the operational reliability of aircraft power plants during flight missions, we analyze the time-varying law and related influencing factors of power plant operational reliability using the machine learning method, meanwhile considering the multi-dimensional and coupling characteristics influencing the operational reliability. An operational reliability analysis method is proposed for power plants considering three factors: the operating state of the power plant, the operating state of the aircraft, and the operating environment of the power plant. Based on the QAR (Quick Access Recorder) data of the actual operation of the aircraft, this method identifies three kinds of factors and 16 main characteristics related to the operational reliability analysis of the power plant. Combined with the space-time relationship of aircraft operation, non-parametric analysis of the working state characteristics and the performance margin of aircraft power plants is conducted using DEA (Data ...

Mode pursuing sampling intelligent exploring method considering expensive con...

LONG Teng, MAO Nengfeng, SHI Renhe, WU Yufei, SHEN Dunliang

doi:10.7527/S1000-6893.2021.25060

2021, (4): 525060-525060.

Abstract

The engineering optimization practices such as modern flight vehicle design often encounter expensive constraints. Based on the standard Mode Pursuing Sampling (MPS) method, a Filter-based Mode Pursuing Sampling intelligent exploring method using Discriminative Coordinate Perturbation (FMPS-DCP) is proposed in this work for constrained optimization problems. In this work, the radial based function network is trained for predicting the values of expansive objective function and constraint functions, and KS function is used to aggregate constraints. Then a filter is constructed for deciding whether to accept sampling points, and a sample point selection strategy is designed to lead the algorithm converge to global feasible optimal value rapidly. FMPS-DCP is tested on a number of standard numerical benchmark problems and compared with CiMPS, Extended ConstrLMSRBF, ARSM-ISES and KRG-CDE. The optimization results indicate that the optimization efficiency of FMPS-DCP is higher than others wi...

Coordination control method for fixed-wing UAV formation through deep reinfor...

XIANG Xiaojia, YAN Chao, WANG Chang, YIN Dong

doi:10.7527/S1000-6893.2020.24009

2021, (4): 524009-524009.

Abstract

Due to the complexity of kinematics and environmental dynamics, controlling a squad of fixed-wing Unmanned Aerial Vehicles (UAVs) remains a challenging problem. Considering the uncertainty of complex and dynamic environments, this paper solves the coordination control problem of UAV formation based on the model-free deep reinforcement learning algorithm. A new action selection strategy, ε-imitation strategy, is proposed by combining the ε-greedy strategy and the imitation strategy to balance the exploration and the exploitation. Based on this strategy, the double Q-learning technique, and the dueling architecture, the ID3QN (Imitative Dueling Double Deep Q-Network) algorithm is developed to boost learning efficiency. The results of the Hardware-In-Loop experiments conducted in a high-fidelity semi-physical simulation system demonstrate the adaptability and practicality of the proposed ID3QN coordinated control algorithm.

Heuristic enhanced reinforcement learning method for large-scale multi-debris...

YANG Jianan, HOU Xiaolei, HU Yu Hen, LIU Yong, PAN Quan, FENG Qian

doi:10.7527/S1000-6893.2020.24354

2021, (4): 524354-524354.

Abstract

Vigorous development of the space industry leads to a nonnegligible space debris threat to future space activities. The Active multi-Debris Removal (ADR) technology has become an indispensable means to alleviate this situation. Aiming at the large-scale multi-debris active removal mission planning problem, a Reinforcement Learning (RL) planning scheme is first proposed based on the maximal-reward optimization model for the ADR problem, and the state, action, and reward function of this problem are defined according to the RL framework. Based on an efficient heuristics method, a specialized Monte Carlo Tree Search (MCTS) algorithm is then presented, with the Monte Carlo Tree Search as the core structure and efficient heuristic operators and reinforcement learning iteration process. Finally, its effectiveness is tested in the large-scale complete Iridium 33 debris cloud. The results show that this method is superior to the original MCTS algorithm and the heuristic greedy algorithm.

Multi-objective optimization algorithm for satellite range scheduling based o...

SUN Gang, CHEN Hao, PENG Shuang, DU Chun, LI Jun

doi:10.7527/S1000-6893.2020.24475

2021, (4): 524475-524475.

Abstract

Progress in China’s aerospace technologies and applications makes the satellite ground station resource shortage problem increasingly prominent. It is necessary to optimize the usage of satellite ground station resources, also called satellite range scheduling problem, which has received extensive attention. Based on the analysis of the characteristics of the problem, the user's preference information on the scheduling results is modeled and formulated, a multi-objective mathematical scheduling model covering the user's preferences is established, and a satellite range scheduling algorithm based on preference-based multi-objective evolutionary algorithm is proposed. To further improve the performance of the proposed algorithm, heuristic strategies based on domain-knowledge including the tasks expansion strategy, conflicts resolution strategy and tasks reduction strategy are designed. Experimental results show that, with the help of user’s preference information, the proposed algorith...

Multi-satellite scheduling approach for emergency scenarios based on hierarch...

LUO Zong, DU Chun, CHEN Hao, PENG Shuang, LI Jun

doi:10.7527/S1000-6893.2020.24721

2021, (4): 524721-524721.

Abstract

Emergency observation mission scheduling is a complex problem of combinatorial optimization with strong timeliness as the scheduling algorithm must complete the computation within the required time limit. Using machine learning methods to provide high-quality initial solutions for scheduling algorithms can effectively simplify the calculation process. For this reason, this paper proposes a multi-satellite scheduling approach for emergency scenarios based on hierarchical forecasting with Transformer network, which decomposes the scheduling into three steps. Firstly, using the Transformer-based task schedulability prediction model to predict whether the observation task will be executed or not, so as to obtain a set of tasks to be executed. After that, using the Transformer-based task allocation model to allocate the satellite to the task set, so as to obtain the initial scheduling scheme. Finally, a constraint modification algorithm based on random hill climbing is used to optimize the ...

Automatic backtracking-based refined segment recognition of maneuvering targe...

QIAO Dianfeng, LIANG Yan, ZHANG Huixia, ZHAO Pengjiao

doi:10.7527/S1000-6893.2020.24744

2021, (4): 524744-524744.

Abstract

Segment recognition of the maneuvering target track is the basis for judging the intention of target's behavior. However, existing track segmentation algorithms have a weak ability to detect changes of pattern, and are thus difficult to meet the requirement of fast and refined track segmentation for maneuvering targets. To this end, our paper proposes a two-layer refined track segmentation framework. The pre-segmentation layer is used to detect the pattern switching during the movement of the target, so as to determine the pre-segment area with obvious pattern changes and obtain the pre-segment points of the area with obvious target pattern changes. Then, iterative backtracking optimization is used to segment the track of the non-pre-segmented area with small differences, so as to obtain more refined segmentation points. The framework has the ability to process the track from coarse to fine segmentation, and can realize the recognition of refined segment of the maneuvering target track...

Adaptive morphological network based UAV target tracking algorithm

LIU Zhenbao, MA Bodi, GAO Honggang, YUAN Jinbiao, JIANG Feihong, ZHANG Junhong, ZHAO Wen

doi:10.7527/S1000-6893.2021.24904

2021, (4): 524904-524904.

Abstract

To solve the problems such as the change of target direction, change of target occlusion and lack of sample diversity in the process of target tracking based on UAV images, this paper proposes a UAV aerial image target tracking algorithm based on the adaptive morphological network. First, the data-driven method is used to expand datasets, and multi rotation angle samples and occlusion samples are added to improve the diversity of samples. The proposed adaptive morphological network improves the deep belief network by rotating invariant constraints to extract deep features with strong representativeness, which enables the model to automatically adapt to the changes of target morphology. The deep feature transformation algorithm is used to obtain the pre-location area of the target to be detected. The target is located adaptively and accurately by the search agent based on the Q-learning algorithm. The category information of the tracking target is extracted by using the deep forest clas...

A pedestrian-aware collision-free following approach for mobile robots based ...

PANG Lei, CAO Zhiqiang, YU Junzhi

doi:10.7527/S1000-6893.2020.24909

2021, (4): 524909-524909.

Abstract

Navigation by following a specified leader is a convenient way for mobile robots. To solve the problem of robot’s following with obstacle avoidance, a collision-free following approach based on path planning is proposed. The static obstacle cost map is generated according to the results of the point cloud segmentation. Multiple disturbing pedestrians are tracked based on the Unscented Kalman Filter (UKF) and Nearest Near Joint Probabilistic Data Association (NN-JPDA) to estimate their motion states, and then the dynamic pedestrian cost map is generated. On this basis, the A*-based global planner plans a global path with the static obstacle cost map, whereas the local planner implemented with Timed-Elastic-Band (TEB) algorithm also takes the dynamic pedestrian cost map into account to plan an optimized local path. This local path tends to follow the global path and helps the mobile robot achieve pedestrian-aware obstacle avoidance. By combining low-frequency global planning with high-f...

ADP-based intelligent attitude fault-tolerant control for launch vehicles

LIANG Xiaohui, HU Changhua, ZHOU Zhijie, WANG Qing

doi:10.7527/S1000-6893.2020.24915

2021, (4): 524915-524915.

Abstract

This paper investigates the problem of attitude control for the launch vehicle with actuator faults. An Adaptive Dynamic Programming (ADP) based intelligent fault-tolerant control scheme is developed, which mainly includes two parts: the stable control part and optimization supplementary part. For the stable control part, the adaptive control technology and the sliding mode variable structure control are used to maintain stability of the fault-tolerant controller, so as to guarantee the stalibilty of the close-loop system and finite-time convergence of the tracking errors. An optimization supplementary controller with the actor-critic structure is also designed to further improve the attitude tracking performance and provide additional compensation control according to the tracking errors caused by actuator faults or external disturbances. The simulation results show that the proposed method can ensure the system stablility and accurate tacking of commands even when there exist malfunctions.

Filter method for dimension reduction in spacecraft autonomous navigation bas...

SUN Bowen, WANG Dayi, WANG Jiongqi, ZHOU Haiyin, GE Dongming, DONG Tianshu

doi:10.7527/S1000-6893.2020.24971

2021, (4): 524971-524971.

Abstract

Motion measurement and state estimation of the non-cooperative target play an important role in the development of space technology. Estimation of the relative state of the non-cooperative target is a difficult problem. In the traditional extended Kalman filter algorithm, it is needed to estimate the centroid position of the non-cooperative target, which increases the dimension of state variables and uncertainty of the system, and thus affects the convergence speed of extended Kalman filtering. In this paper, a relative navigation method for non-cooperative target based on the sequence image is proposed. The attitude estimation of non-cooperative target can be realized without estimation of the centroid position, then the centroid position can be estimated based on the attitude estimated before. The relationship between the measured value and the true attitude of the non-cooperative target is derived. The sequence-image based measurement model is constructed. The state formula without ...

Fault tolerant game control of swarm confrontation with decision faults

NI Yuan, YANG Hao, JIANG Bin

doi:10.7527/S1000-6893.2020.24978

2021, (4): 524978-524978.

Abstract

Some UAVs of the swarm are attacked or manipulated by the enemy in the confrontation scenario, and their decision-making rules are tampered to lead to a deviation between the swarm behaviors and the expected equilibrium. In the framework of the evolutionary matrix game, the equations for the multi- population replicator dynamics are used to model the UAV swarm and faults. Based on the Lyapunov functional approach, the local asymptotic stability of the equilibrium point as well as its domain of attraction is analyzed in both healthy and faulty cases. A self fault-tolerant condition is established. Moreover, an incentive based cooperative fault tolerant game control method between clusters is proposed to compensate for the deviation of swarm behaviors. With the method, the task allocation state of the UAV swarm can still reach the desired equilibrium point when there are faults, and the ideal payoff of the division of labor can be obtained.

Relative pose estimation and optimization of a failure satellite with low-lig...

LIU Fucheng, MU Jinzhen, LIU Zongming, HAN Fei, LI Shuang

doi:10.7527/S1000-6893.2021.24984

2021, (4): 524984-524984.

Abstract

The serious damage of information in images under a low-light imaging condition may reduce the accuracy and robustness of the pose estimation of a failure satellite. Thus, this paper proposes an unsupervised Generative Adversarial Network (GAN)-based low-light image enhancement model. The generator is based on the U-Net with dense residual connection. The discriminator is designed as a global-local structure, and the traditional single scalar output of the discriminator is extended to the multiple scalar output. The single natural image Generative Adversarial Network (SinGAN) is improved based on evolutionary and parallel training methods to augment the few-shot training samples. Finally, a local map of feature information is established based on initialization of ORB-SLAM to alleviate dependence of pose estimation on reference frame. Through dense matching of Region of Interest (ROI) in key-frames, a nonlinear optimization model of the normal vector of plane and the mounting height of...

An anomaly detection method for irregularly sampled spacecraft time series da...

YAN Tijin, XIA Yuanqing, ZHANG Hongwei, WEI Minfeng, ZHOU Tong

doi:10.7527/S1000-6893.2021.25019

2021, (4): 525019-525019.

Abstract

Real time anomaly detection of spacecraft telemetry data is critical for space mission. Previous methods mostly model the regularly sampled time series data with low missing rates. However, spacecraft telemetry time series data has the characteristics of large dimensions, many noises, high missing rates, irregular sampling intervals, and it is thus more difficult to conduct anomaly detection tasks. An Irregularly sampled Multivariate time series Anomaly Detection (IMAD) algorithm is proposed to model the irregularly sampled multi-dimensional spacecraft time series data with missing values. First, Gated Recurrent Unit with trainable Decays (GRU-D) to model missing values and irregular sampling intervals. Furthermore, the variational autoencoder is used to model the randomness and learn the distribution of normal time series data. Finally, the adaptive threshold determination method based on the extreme value theory is adopted to determine the appropriate threshold for anomaly detection....